Five silicon qubits share a line—and still hit 99.99%

Picture five tiny electron spins ticking like metronomes inside a sliver of silicon—then imagine trying to nudge each one with microwaves, all at once, without throwing the others off. That’s the control problem that kept multi‑qubit silicon chips from behaving as well together as they do alone.

In a Nature Physics study, researchers at the RIKEN Center for Emergent Matter Science and collaborators reported that they kept several silicon spin qubits running in parallel while maintaining extremely high accuracy. They worked with an array of five qubits fed by a shared microwave control line and still pushed single‑qubit gate errors down to parts per ten‑thousand—and in one case, parts per hundred‑thousand.

Why this is in the news today

This week’s result lands because parallelism is the roadblock between lab‑scale demonstrations and useful quantum processors. The team reports highly accurate control of several silicon spin qubits operating simultaneously in one chip‑level setup, a capability that underpins any realistic plan to scale silicon devices using shared control hardware.

The intuitive picture, and where it breaks

If you’ve tuned a radio while someone else plays guitar nearby, you know signals can overlap and cause detuning or buzz. It’s tempting to think qubits just “hear” each other’s microwaves the same way. The analogy breaks down because a qubit’s quantum phase—how its wavefunction’s peaks and troughs line up—matters as much as its frequency, and microwave control can shift that phase in subtle, cumulative ways.

The hidden culprit: AC Stark shifts

The team traced a big share of the simultaneous‑operation error to the AC Stark shift: when a microwave drive slightly alters a qubit’s effective energy splitting, its state picks up an unintended phase. In isolation, you calibrate and cancel that. With neighbors driven at the same time, cross‑couplings make those shifts add up unpredictably unless you account for them.

In the study’s five‑qubit array on a shared microwave line, those shifts were the mechanism that quietly degraded accuracy when several qubits were active together. Identifying that specific effect let the researchers target it rather than treating “parallel control” as a vague source of noise.

A scalable fix: pairwise calibration on a shared line

Calibrating every possible multi‑qubit combination would explode in complexity as chips grow. Instead, the researchers implemented a scheme that relies on calibrations performed between pairs of qubits. By measuring and compensating the AC Stark‑induced phase shifts pair by pair, they could predict and correct the composite shifts when larger subsets ran together.

Crucially, the array used a shared microwave control line. That’s a design you can actually lay out across bigger chips, because giving every qubit its own fully independent control hardware doesn’t scale gracefully. The results show that shared control doesn’t have to mean sloppy control—if you calibrate the right thing.

Operating condition	Reported performance
Single qubit tested alone	Each achieved single‑qubit gate fidelity well above 99.99% [c3]
Best single qubit	Highest fidelity approaching 99.999% [c4]
Up to three qubits run simultaneously	Fidelity remained above 99.99% [c5]
All five qubits run simultaneously	Accuracy stayed at approximately 99.9% [c6]

What still fails when you fix the phase

Even after correcting AC Stark shifts, the last sliver of accuracy didn’t vanish. The team linked the remaining loss to drive‑induced decoherence: when you turn up microwave power to steer several qubits at once, you also accelerate processes that sap quantum coherence. That’s not a miscalibrated phase; it’s the quantum state itself losing definition.

Further improvements to device design could mitigate this source of error, according to Leon Camenzind. That points the next round of work toward hardware—materials, layouts, filtering—rather than only smarter pulse shaping.

Why silicon and why shared lines

Silicon spin qubits are built in the same material family as everyday chips, so arrays and control wiring are natural fits for semiconductor processes. That makes the question of wiring density and routability—like whether you can share control lines—more than an academic point; it’s a practical constraint on how many qubits you can fit and actually drive.

By showing highly accurate control with a shared microwave line across five qubits, the study supports architectures that avoid per‑qubit bespoke control. That’s one path toward the stated long‑term goal, from Seigo Tarucha, of building a fault‑tolerant quantum computer based on silicon spin qubits.

What the study doesn’t tell us (yet)

The report gives headline fidelities and identifies AC Stark shifts and drive‑induced decoherence, but it does not specify the detailed hardware geometry or the exact pulse families used to achieve those numbers. It also doesn’t enumerate how the pairwise calibration overhead scales in wall‑clock time on much larger arrays. Those omissions don’t undercut the core result; they just mark where replication studies and engineering follow‑ups will need to fill in specifics.

The road from five to fault‑tolerant

The researchers’ technique shows that accuracy above 99.99% is compatible with running several qubits at once on a shared line, and that even all five together held at about 99.9%. That’s a concrete step toward parallel, scalable control rather than isolated demonstrations.

The team links the residual errors to mechanisms that have engineering handles—calibration for phase shifts, device design for drive‑induced decoherence. As those handles are turned, the long‑term goal, stated by Seigo Tarucha, is a fault‑tolerant quantum computer built from silicon spin qubits.